The structural challenge Ajunwa's blueprint has to solve before anything else is the talent pipeline. Every previous technology-specific federal commission was eventually captured by the industry it regulated because the people qualified to do the regulating came from the same labour pool as the people being regulated. The FCC and telecoms. The FAA and aviation. The revolving door is a structural feature of specialist regulation, not a corruption problem.
For AI the pipeline problem is worse because the expertise is newer and more concentrated. The population who understand frontier model development deeply enough to write meaningful regulation might be measured in hundreds, and most of them currently work at the same five companies. An AI commission staffed by people who dont understand the technology produces compliance theatre. One staffed by people who do produces regulatory capture. Navigating between those two failure modes is the actual governance problem, and its harder than the policy design.
Governance conversations like this matter — and I'd keep one thread visible: AI has no independent compass, it amplifies the one its operators bring. Commissions and policy are necessary but downstream; the first regulation is the self-knowledge of the people wielding it. Free book on that thru 6/3: amazon.com/dp/B0H3HY8W9F
The structural challenge Ajunwa's blueprint has to solve before anything else is the talent pipeline. Every previous technology-specific federal commission was eventually captured by the industry it regulated because the people qualified to do the regulating came from the same labour pool as the people being regulated. The FCC and telecoms. The FAA and aviation. The revolving door is a structural feature of specialist regulation, not a corruption problem.
For AI the pipeline problem is worse because the expertise is newer and more concentrated. The population who understand frontier model development deeply enough to write meaningful regulation might be measured in hundreds, and most of them currently work at the same five companies. An AI commission staffed by people who dont understand the technology produces compliance theatre. One staffed by people who do produces regulatory capture. Navigating between those two failure modes is the actual governance problem, and its harder than the policy design.
Governance conversations like this matter — and I'd keep one thread visible: AI has no independent compass, it amplifies the one its operators bring. Commissions and policy are necessary but downstream; the first regulation is the self-knowledge of the people wielding it. Free book on that thru 6/3: amazon.com/dp/B0H3HY8W9F
I proposed a similar commission to address what I saw as the risks posed by digital rights management back in 2004. Sadly, like that proposal, I don’t see this as politically viable https://lawreview.colorado.edu/wp-content/uploads/2025/12/Woodford.pdf